GSoC 23 | Databases | Interested in formal methods https://xitter.cf/t.co/5b1BzaKFVL

Joined October 2021
Was trying to update sqlite version for Turso and add more tests, and somehow ended up finding 3 optimizer bugs in SQLite. Wasn’t expecting that, but hey, I’ll take it.
3
4
3
45
20,335
I feel like i shouldn't have access to llm's, i'm learning distributed systems - by reading code most of the times and just writing spec in quint by hand. but when something doesnt click i find myself reaching out to gpt and i feel more retarded not more understood.
6
27
541
i mean ofc i am retarded before this too but i fucking hate how easily accessible this is.
4
49
why can't we just do masters/graduate program without finishing undergrad 🤡
4
57
Pavan-Nambi retweeted
I wrote a new blog post about MVCC in @tursodatabase . The post does not focus on one thing in the system but instead follows the life of a row. From birth up to ultimately its death turso.tech/blog/mvcc-row-lif….
7
2
60
4,519
Pavan-Nambi retweeted
Our product accidentally found a bug in Postgres! I always hold my company to the highest standards when it comes to software verification. One of the internal metrics we track is how often we find rare correctness issues in some of the world's most well-tested systems. Some of the customers we work with are building things where any unexpected failure could be catastrophic. That often means we verify all the upstream dependencies their products rely on in our environments, too. Sometimes, that turns up interesting findings like this one: while verifying a customer’s software, our simulation engine found a rare bug in Postgres.
16
8
4
88
15,347
Pavan-Nambi retweeted
What are the best books on formal verification I should read? Asking for a friend…
Went back and finished my Viewstamped Replication implementation in Rust. Protocol only, I/O is provided by the caller. Verified with a TigerBeetle-style simulator. Now thinking about doing some verification with Lean... github.com/penberg/vsr-rs
11
11
1
151
16,211
Pavan-Nambi retweeted
“Out in the wild, most systems are very un-specifiable, and that’s a big problem if we want to formally verify a lot more of the world.” Pretty much sums up my skepticism about people (and companies) saying we’ll verify all software. galois.com/articles/specific…
5
7
3
56
9,826
I suffer with this constantly 😭
AI allows cheap, rapid coding experiments. the worst thing you can do is leave the experiment in a state where you didn't abandon it but didn't fully see it through to production either. commit to one of those two or your experiments become a disguised form of procrastination
2
182
Pavan-Nambi retweeted
James and I with @tursodatabase CEO @glcst proving that you can't put a bunch of database founders in a room together without a hard time limit and/or content filter.
What happens when you put three database founders on a podcast? Surprisingly, no fights! New episode of Databased is out now. Full video in the thread 👇
3
3
44
5,068
Software infrastructure at a glance (August 28, 2026) @ScyllaDB hosts P99 Conf on October 21. @tursodatabase writes about Turso's MVCC. @ArgaLabs raises a $10M seed round. And @mozilla is hiring a Senior Staff Performance Engineer, Firefox remotely from Canada.
2
5
525
Pavan-Nambi retweeted
Got my first paper award!
25
14
1
311
18,234
This is like the subprime of databases… But most (still) don’t recover the WAL correctly in at least two ways cf.: 1. “Can Applications Recover from fsync Failures?” 2. “Protocol-Aware Recovery for Consensus-Based Storage” Single node DBs like SQLite can lose committed transactions on storage faults. And distributed protocols (and DBs built on them) like Paxos and Raft can enter split brain and lose the cluster. The papers are worth a slow and careful read.
Yes, that was the fix that most DBMSes applied for fsyncgate But most still don’t use Direct I/O when they replay the WAL on reboot. So they risk externalizing non-durable data as durable (because they still read from the incoherent kernel page cache). cf. “Can Applications Recover from fsync Failure?” which looked into this in 2020, two years after these fixes, and found that PG, SQLite et al were still vulnerable.
4
23
3
142
15,667
Pavan-Nambi retweeted
Handling edge-case failures like this is way simpler in distributed storage databases (e.g. Aurora) and distributed databases (e.g. Aurora DSQL) than in single-system databases. It's one of many cases where distribution simplifies system properties. Here's how it works in DSQL:
Does anyone know good blogs around how databases handles fsync failures?
5
16
2
234
27,615
Pavan-Nambi retweeted
On some ICFP related news, extended version of our ICFP23 article "Etna: An Evaluation Platform for Property-Based Testing" is accepted to Journal of Functional Programming; and as of today, "Programmable Property-Based Testing" is accepted for publication in ICFP26!
4
1
89
23,000
Does anyone know good blogs around how databases handles fsync failures?
7
2
1
27
27,407
should* handle
553
Sep 21: @doris_apache Sep 28: @OneWillAI Oct 5: @sentry Oct 19: FloeDB Oct 26: @databricks Lakebase Nov 2: @ClickHouseDB PostgreSQL Nov 9: @daftengine Multibase Nov 16: @tursodatabase Nov 23: @DB9AI Nov 30: @Teradata
9
2
84
5,990
Pavan-Nambi retweeted
Not sure how helpful this is, but any devs having a problem in the current job market... Do everything you can to deeply learn Database internals. There are so many jobs for DB engineers right now, I swear I get at least 1 email every week or so from some company looking for someone to work on storage engines/query engines/filesystems/search/indexing type of stuff. you will not regret databasemaxxing 🙏 pthorpe92.dev/databasemaxxin…
13
38
3
525
59,248